CUDA Runtime benchmark. # Download dataset Please download the dataset contained in [Rodinia 3.1](http://lava.cs.virginia.edu/Rodinia/download_links.htm). Copy the data directory into rodinia. # Environment variables Set the environment variables in scripts/environments, including dataset path, AvA modes, etc. The library paths in `opencl/util/make.mk` and `cuda/util/nvidia.mk` need to set appropriately. # Run benchmarks In `scripts/`, `test_cl.sh` is used to run all OpenCL benchmarks and `test_cu.sh` is for CUDA benchmarks. `parse.sh` is used to compute the average end-to-end execution time. Only `test_cudart.sh` is supoorted in this repository. # Run individual benchmarks Every benchmark has a `run` bash script in their corresponding directory. Simply execute the script to run the benchmark.